The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill accesses local files to retrieve authentication tokens for external services.
Evidence: The script scripts/welcome_client_emails.py loads token.json from the filesystem to authenticate the Google API client. The SKILL.md file also explicitly lists credentials.json as the source for authentication credentials.
[INDIRECT_PROMPT_INJECTION]: The skill incorporates external client information into email content without sanitization, which creates a surface for manipulating agent or recipient behavior via untrusted data.
Ingestion points: The client_name and company_name fields in scripts/welcome_client_emails.py ingest data directly from the input payload.
Boundary markers: None; the skill does not use delimiters or instructions to isolate the untrusted data from the email templates.
Capability inventory: The skill is capable of sending automated emails via the googleapiclient and Gmail API.
Sanitization: No input validation, escaping, or filtering is performed on the data before it is interpolated into the HTML email bodies.

welcome-email